import numpy as np
import pandas as pd
import holoviews as hv
import bokeh
hv.extension('bokeh')
import seaborn as sns

%load_ext autoreload
%autoreload 2

import erg

ergram = erg.ERG(r'../data/BYB_Recording_2020-12-23_13.47.39.wav')

Data validation: visualize TTL pulses for each frequency

For a specific frequency, this is where all of that frequency occurred for each color.

hv.Curve(
    data=ergram.df,
    kdims=['time (s)'],
    vdims=['TTL', 'theoretical frequency','color'],
).groupby(
    ['color','theoretical frequency']
).overlay(
    'color'
).layout(
    'theoretical frequency'
)

Hey ET! You can zoom in on each data point using the HoloViews explorer. This is just a way to look at where each color occurs in the experiment for each frequency, tee-ing up a similar graph below.

Visualize responses to UV light

%%opts Curve {+axiswise}
# per https://github.com/holoviz/holoviews/issues/2910, (alternatively, https://github.com/holoviz/holoviews/issues/2343),

hv.Curve(
    data=ergram.df[ergram.df.color=='UV'],
    kdims=['time (s)'],
    vdims=['channel 1', ('theoretical frequency', 'frequency (Hz)')],
).groupby(
    ['theoretical frequency']
).layout(
    'theoretical frequency'
)

So this is kind of interesting. Toward the end of the end of stimulation, especially with prolonged periods of stimulation (e.g., at high frequencies or low frequencies), you get some movement artifact. I saw the cockroaches really seem to not like that- they would squirm a lot. Another observation which I think is even more important, at around the halfway point, you see some short upward notch.

Visualize responses to blue light

%%opts Curve {+axiswise}
# per https://github.com/holoviz/holoviews/issues/2910, (alternatively, https://github.com/holoviz/holoviews/issues/2343),

hv.Curve(
    data=ergram.df[ergram.df.color=='B'],
    kdims=['time (s)'],
    vdims=['channel 1', ('theoretical frequency', 'frequency (Hz)')],
).groupby(
    ['theoretical frequency']
).layout(
    'theoretical frequency'
)

Visualize responses to green light

%%opts Curve {+axiswise}
# per https://github.com/holoviz/holoviews/issues/2910, (alternatively, https://github.com/holoviz/holoviews/issues/2343),

hv.Curve(
    data=ergram.df[ergram.df.color=='G'],
    kdims=['time (s)'],
    vdims=['channel 1', ('theoretical frequency', 'frequency (Hz)')],
).groupby(
    ['theoretical frequency']
).layout(
    'theoretical frequency'
)

Visualize responses to red light

%%opts Curve {+axiswise}
# per https://github.com/holoviz/holoviews/issues/2910, (alternatively, https://github.com/holoviz/holoviews/issues/2343),

hv.Curve(
    data=ergram.df[ergram.df.color=='R'],
    kdims=['time (s)'],
    vdims=['channel 1', ('theoretical frequency', 'frequency (Hz)')],
).groupby(
    ['theoretical frequency']
).layout(
    'theoretical frequency'
)

Visualize responses to IR light

%%opts Curve {+axiswise}
# per https://github.com/holoviz/holoviews/issues/2910, (alternatively, https://github.com/holoviz/holoviews/issues/2343),

hv.Curve(
    data=ergram.df[ergram.df.color=='IR'],
    kdims=['time (s)'],
    vdims=['channel 1', ('theoretical frequency', 'frequency (Hz)')],
).groupby(
    ['theoretical frequency']
).layout(
    'theoretical frequency'
)

I know this is lot of stuff. I wanted to explicitly spell out how cockroaches respond to increasingly long wavelengths. Note how the y-axis changes with color: red responses are actually quite low, and IR are nonexistant.

TODO:

• Visualize the peak of each stimulation
• Model assessment of exponential vs null hypothesis for decay? (Has this question been asked before? Why would it be exponential as opposed to something else?)